Treball Final de Graudiposit.ub.edu/dspace/bitstream/2445/175436/1/TFG_QU...uno de los problemas....

Tutor/s

Dr. Oscar Núñez Burcio Departament d’ Enginyeria Química i

Química Analítica

Dr. Javier Saurina Purroy Departament d’ Enginyeria Química i

Química Analítica

Treball Final de Grau

HPLC-FLD and HPLC-UV Fingerprinting for the Characterization, Classification and Authentication of Paprika.

Caracterización, Clasificación y Autenticación de Pimentones mediante huellas HPLC-FLD y HPLC-UV.

Luis R. Rodríguez Javier January 2021

Aquesta obra esta subjecta a la llicència de: Reconeixement–NoComercial-SenseObraDerivada

http://creativecommons.org/licenses/by-nc-nd/3.0/es/

No human investigation can be called real knowledge if it does not pass through

mathematical demonstrations.

Leonardo da Vinci

I would like to thank my supervisors in this work, Oscar Núñez and Javier Saurina, that have

guided and helped me during the whole project while giving me the space to grow as a chemist,

allowing me to improve my liquid chromatography and chemometrics skills and knowledge.

I would also like to thank all the members of CECEM for accepting me with open arms,

especially to Guillem, Nerea, and Josep, for being there when I needed them the most.

I can’t forget my friends, family, and college classmates, who have made me who I am

today. In particular, I would like to thank Aïda Valverde and Victor Rubio, who have

accompanied me during these months, motivating me to move on, not only in the project but

also in the day-to-day.

REPORT

HPLC-FLD and HPLC-UV Fingerprinting for the Characterization, Classification and Authentication of Paprika. 1

CONTENTS

1. SUMMARY 3

2. RESUMEN 5

3. INTRODUCTION 7

3.1. Food authentication 7

3.2. Paprika 8

3.3. Analytical methodologies for food fraud detection 9

3.3.1. Targeted analysis 10

3.3.2. Non-targeted analysis 10

3.4. Chemometrics 11

3.4.1. Principal component analysis 11

3.5.2. Partial least squares - discriminant analysis 12

4. OBJECTIVES 13

5. EXPERIMENTAL SECTION 15

5.1. Reagents and solutions 15

5.2. Instrumentation and chromatographic conditions 15

5.3. Samples and sample treatment 16

5.4. Data analysis 17

6. RESULTS AND DISCUSSIONS 19

6.1. Chromatographic separation 19

6.2. Chromatographic fingerprints 20

6.2.1. HPLC-FLD chromatographic fingerprints 21

6.2.2. HPLC-UV chromatographic fingerprints 22

6.3. Exploratory studies by PCA 23

6.3.1. HPLC-FLD study by PCA 23

6.3.2. HPLC-UV study by PCA 24

6.4. Classification of samples by PLS-DA 25

2 Rodríguez Javier, Luis R.

6.4.1. HPLC-FLD classification by PLS-DA 26

6.4.2. HPLC-UV classification by PLS-DA 29

6.5. Validation results 30

7. CONCLUSIONS 33

8. REFERENCES 35

9. ACRONYMS 39


1. SUMMARY

In recent years, there has been a growing interest in the quality of natural foods from

society. Consequently, the demand for food traceability has increased and, along with it,

measures to satisfy consumer concerns. One of these tools is the Protected Designation of

Origin (PDO), which guarantees the quality and geographical origin of a product. In the case of

paprika, a red spice obtained after drying and grinding certain varieties of red peppers, there are

seven PDOs in the European Union. Even with this quality seal, the products do not escape

from food fraud like adulterations with chemicals or paprika from other geographical origins. To

avoid this, it is essential to develop new techniques to solve these problems.

This project aims to create a non-targeted method that allows the classification of paprika

according to their production region. To this end, chromatographic fingerprints obtained by high-

performance liquid chromatography with fluorescence (HPLC-FLD) and ultraviolet (HPLC-UV)

detection were proposed as chemical descriptors. The chromatographic separation was

performed on a C18 reversed-phase column under gradient elution using acidified water (0.1%

formic acid) and acetonitrile as the mobile phase components.

The obtained chromatographic fingerprints were analyzed using chemometric techniques

such as principal component analysis (PCA) and partial least squares - discriminant analysis

(PLS-DA). In these studies, apart from ensuring that the proposed methods were robust and

reproducible, a classification of the samples with 95.8% accuracy was obtained for both HPLC-

UV and HPLC-FLD fingerprints.

Keywords: Paprika, Food fraud, PDO, fingerprints, HPLC-FLD, HPLC-UV, PCA, PLS-DA.


2. RESUMEN

En los últimos años, ha habido un interés creciente en la calidad de los alimentos naturales

por parte de la sociedad. En consecuencia, ha aumentado la demanda de la trazabilidad

alimentaria y, junto a ella, medidas para saciar la preocupación de los consumidores. Una de

estas herramientas es la designación de Denominación de Origen Protegido (DOP), la cual

permite garantizar la calidad y el origen geográfico de un producto. En el caso del pimentón,

una especia roja obtenida tras secar y triturar ciertas variedades de pimientos, se encuentran

siete DOP en la Unión Europea. Aún con este sello de calidad, los productos no se libran del

fraude alimentario, llegando a encontrar adulteraciones con químicos o alimentos de otras

regiones. Para evitarlo, es importante el desarrollo de nuevas técnicas que se adapten a cada

uno de los problemas.

Este proyecto tiene como objetivo la creación de un método no dirigido que permita la

clasificación de pimentones en función de su región de producción. Con este fin, se han

propuesto como descriptores químicos las huellas cromatográficas obtenidas mediante

cromatografía líquida de alta eficacia con detección de fluorescencia (HPLC-FLD) y ultravioleta

(HPLC-UV). La separación cromatográfica se ha llevado a cabo con una columna de fase

invertida C18 con gradiente de elución utilizando agua acidificada (0.1% de ácido fórmico) y

acetonitrilo como fase móvil.

Las huellas cromatográficas obtenidas han sido analizadas empleando técnicas

quimiométricas como el análisis de componentes principales (PCA) y el análisis discriminante

por regresión de mínimos cuadrados parciales (PLS-DA). En estos estudios, aparte de

confirmar que los métodos propuestos son robustos y reproducibles, se ha obtenido una

clasificación de las muestras con un 95.8% de éxito para ambas huellas HPLC-UV y HPLC-

FLD.

Palabras clave: Pimentón, Fraude alimentario, DOP, Huella cromatográfica, HPLC-FLD,

HPLC-UV, PCA, PLS-DA.


3. INTRODUCTION

3.1. FOOD AUTHENTICATION

In the last years, the consumption of natural products instead of processed ones has

increased significantly. Even though there has never been so much interest in healthy food as

there is today, consumers still increase the demand about information, either to avoid certain

allergens, to have a balanced diet, or to know the origin [1].

The public awareness about food safety and quality causes the constant development in

food authentication, which is the process of verifying that the product is in compliance with the

label description, including among others, the origin (geographical, species or genetic),

production method, and processing technologies [2].

One of the biggest problems in this situation is food fraud, defined as a term that

encompasses the deliberate addition, substitution, misrepresentation or tampering of a product

for economic benefit at the expense of the consumer [3]. Natural products are prone to be

adulterated with more economic additives, and these adulterants range from natural products

with similar properties to harmful chemicals. Some examples are the adulteration of extra virgin

olive oil with refined oils [4], Sudan dyes in spices [5] or the melamine adulteration of milk

powder [6]. Although they are not equated, these previous situations are considered a public

health risk and, for that reason, it is important that the food industry has a good traceability,

quality control, and security to ensure that the products satisfy consumer expectations.

In order to control these measures over the world, one of the many options are the

regulations made by the International Standardization Organization (ISO) such as ISO:9001

(quality management) or ISO:22005:2007 (traceability in the food chain). Furthermore, the

European Union has its own institution for food legislation. The European Food Safety Authority

(EFSA) includes a set of laws and guidelines to assure food quality and safety [7]. With the aim

of controlling the product precedence as well as to support the good practices in rural and

agricultural activities, the Council Regulation (EC) No 510/2006 [8] proposes, among others, the

Protected Designation of Origin (PDO), which relates the products to the geographical area


where they are prepared, processed and produced. Even so, these products are not spared

from food fraud and authentication has an important role preventing it.

3.2. PAPRIKA

Paprika, or chili pepper, is a red powder spice obtained after drying and grinding some

varieties of red peppers of the genus Capsicum, which belongs to the Solanaceae family [9].

Due to its characteristic color and taste, paprika is commonly used to add flavor and color to

various ethnic dishes and many foods such as soup, beverages, drinks, meat, baked goods,

candy, ice cream, and seasoning mixes [10], although others applications as cosmetics,

personal protection sprays, adsorbents to remove contaminants, and medicine are reported

[11,12].

The pepper used for the preparation of paprika has its origin in the Mexican Indians’ diet,

with approximately 39 wild, semidomestic, and domestic species within the genus Capsicum,

the most common is the Capsicum annuum [11,13]. With the marine exploration period at the

end of the XV century, these seeds arrived to Europe and their farming was adapted to the

conditions of each region. Since then, the globalization of paprika increased exponentially due

the ease of hybridization [11]. Nowadays, C. annuum is one of the most cultivated vegetables in

the world [14].

In Europe seven Protected Designation of Origin for paprika exist: Slovakia Žitavská paprika

(Slovakia), Kalocsai fűszerpaprika-őrlemény and Szegedi fűszerpaprika-őrlemény (Hungary),

Piment d’Espelette (France), Pimentón de la Vera (Extremadura, Spain), Pimentón de Murcia

(Murcia, Spain), and since 2020, Pebre bord de Mallorca (Mallorca, Spain) [15]. It is obvious to

think that the paprika from two regions would taste different due to natural and human factors

such as the climatic condition, cultivation area and techniques, water resources, soil

management or manufacturing practices, but properties and organic compounds of paprika are

also specific for each region [16].

Regarding the flavor variety, three important groups can be distinguished: hot paprika,

sweet paprika, and bittersweet paprika. Besides, during the elaboration of the paprika, in some

cases peppers are dried by employing oak wood smoke, adding three new important flavors:

smoked hot paprika, smoked sweet paprika, and smoked bittersweet paprika [17].


Concerning the organic compounds present in paprika, flavonoids and carotenoids are

responsible for the characteristic red color [11] while capsaicinoids and capsinoids grant the

pungency (spicy flavor) and some health benefits along with other bioactive compounds [18].

Some examples are ascorbic acid (vitamin C) or tocopherols (vitamin E), but among the most

important ones are phenolic and polyphenolic compounds, due to their antioxidant activity that

can promote vascular protection, antiosteoporotic, anti-inflammatory, antitumor, antiobesity,

antiallergic and analgesic effects amidst others [12,16,19].

Apart from its great utility in health, with the advance of the analytical methodologies,

polyphenol content and distribution have been used for the evaluation of food quality and the

detection of frauds in paprika [18].

3.3. ANALYTICAL METHODOLOGIES FOR FOOD FRAUD DETECTION

With the increasing food fraud complexity, an improvement in the analytical approaches is

required. In food authentication there are several methodologies based on infrared, Raman, and

nuclear magnetic resonance (NMR) spectroscopies, isotopic analysis by nuclear magnetic

resonance, chromatographic techniques, fluorescent spectroscopy, ultraviolet-visible (UV-VIS)

spectroscopy, and DNA-based technology among others [20]. Nowadays, liquid

chromatography with ultraviolet detection (LC-UV) or coupled to mass spectrometry (LC-MS) or

high-resolution mass spectrometry (LC-HRMS) are some of the most useful techniques to

detect fraud [21].

Regarding paprika, studies with UV-VIS [22], visible - near infrared [23], Fourier transform

mid-infrared [24], Raman [25], NMR [26], and energy dispersive X-ray fluorescence [15]

spectroscopies, as well as liquid chromatography with electrochemical detection using screen-

printed carbon-based nanomaterial electrodes [9], LC-UV [27], LC-MS [12], and LC-HRMS [13]

have been employed to assess their characterization, classification and authentication.

As seen in these works, regardless the instrumentation, two analytical strategies can be

mainly employed: targeted and non-targeted methodologies. The percentage of food

authentication works using those approaches has increased by 300% between 2007 and 2016

due the rising demand in our society to guarantee food authenticity [28].


3.3.1. Targeted analysis

Targeted analysis is an analytical strategy whose objective is the detection or quantification

of pre-defined target compounds. When these targets are linked to the authentication issue,

they are called analytical markers and two types can be distinguished. Primary markers refer to

those whose result is directly related to the authentication, often based on specific legal limits.

Secondary markers are those indirectly related, for example, those associated with the

geographic origin, specie or agricultural production methods [28]. In the case of paprika these

markers usually are polyphenols or capsaicinoids [13].

Targeted methods are often quantitative and have a greater sensitivity and selectivity

respect to non-targeted methods, but, as only one or a few analytes are studied, the information

offered to protect the consumer against adulteration may be limited. This strategy is usually

used when the suspected target is a primary marker, otherwise non-targeted analysis is

preferably employed.

3.3.2. Non-targeted analysis

Non-targeted analysis, also known as untargeted analysis or fingerprinting, relies on

complex instrumental data that can be generated with a wide variety of analytical methodologies

whose objective is to obtain as many features (peaks or signals) of metabolites as possible

without the requirement of identifying the compounds [29]. These methods usually are

qualitative and no markers are defined.

Food discrimination and classification are carried out contrasting and comparing patterns

between the obtained fingerprints. This pattern represents the metabolites that change in

response to natural or artificial alterations, for example, geographical origin or adulteration [29].

In order to avoid a loss of signals or information, it is important to use sample treatments as

simple and unspecific as possible [30]. The ability to detect multiple small variations in the food

product makes untargeted methods a perfect option to deal with the complexity of modern food

fraud, especially when no primary or secondary markers are available [28].

The principal disadvantage of these methodologies arises in the large amount of chemical

data obtained. First, a data matrix is built converting the fingerprints into valuable information

with a specific software. Then, all this data is analyzed with a multivariate statistical program,

so, in order to prevent food fraud, a chemometric study has to be done.


3.4. CHEMOMETRICS

Chemometrics is defined by the International Chemometrics Society as “the chemical

discipline that uses mathematical, statistical, and other methods employing formal logic to

design or select optimal measurement procedures and experiments, and to provide maximum

relevant chemical information by analyzing large amounts of chemical data”. Many applications

of chemometrics have been proposed in the last decade to solve the adulteration and food fraud

problematic due the advantages of multivariate statistical methods [31].

One of the main classifications in chemometric distinguishes between supervised and non-

supervised methods. Non-supervised methods consist of identifying groups or connections

between samples without any prior knowledge of the class or group to which they belong.

Instead, supervised methods require previous information, such as the belonging class, and

uses it as a reference to predict the results. Two representative techniques of these methods

are principal component analysis (PCA) and partial least squares - discriminant analysis (PLS-

DA), respectively.

3.4.1. Principal component analysis

Principal component analysis is a non-supervised multivariate technique that decomposes

the data in a set of orthogonal variables called principal components (PCs). These PCs reduce

the original data dimension and only contain the most valuable information. Each PC contains a

determined variance of the original variables, the first PC (PC1) describes the maximum amount

of the data variance, and the subsequent (PC2, PC3…) provide progressively less variance

[31].

The representation of the sample coordinates of two PCs can be displayed in a two-

dimension diagram named scores plot. Samples with similar properties shape clusters and the

greater the difference, the greater the separation between them. Furthermore, a projection of

the variables of the PCs can be employed to show how strongly each characteristic influences a

PC. This representation is called loading plot and allows to obtain information of the most

descriptive or discriminant variables.

A PCA model provides an exploratory analysis of all the information in a data table, offering

an overview of the groupings, trends, patterns, and outliers of the studied samples [32]. For that

reason, PCA is used in numerous applications as a first step [31].


3.4.2. Partial least squares - discriminant analysis

Partial least squares - discriminant analysis is a supervised classification method that

relates the experimental data (here fingerprints, as the chromatograms of the samples) with

their class membership. In PLS-DA converts the original data into linearly uncorrelated variables

called latent variables (LVs) expressing the quantitative correlation between an X matrix (data

fingerprints) and a Y matrix (class belonging).

This method seeks for the maximum covariance between these matrices and for that, a

confusion matrix is generated with the proposed classification models. Then, a prediction of an

“unknow set” is carried out in order to check the model performance, i.e. how many samples are

correctly classified [31].

In order to obtain the best prediction in the classification - validation, the optimal number of

LVs has to be chosen. Too few may retain insufficient information in the calibration, and too

many may lead to overfitting so that the model is not robust in front of small variations. For that

reason, the ideal number of LVs is the one with the minimum error in the model approach.


4. OBJECTIVES

This project aims to develop non-targeted methods based on high-performance liquid

chromatography with fluorescence (HPLC-FLD) and ultraviolet (HPLC-UV) detection to achieve

the characterization, classification, and authentication of paprika samples regarding their

geographical indication. The next steps will be performed to achieve this objective:

1. Paprika samples will be submitted to a simple treatment to extract their bioactive

compounds.

2. The gradient elution used in the chromatographic separation will be optimized

based on a paprika sample analyzed by the HPLC-FLD method.

3. The obtained extracts will be analyzed employing HPLC-FLD and HPLC-UV to

obtain the sample chromatographic fingerprints.

4. PCA will be applied to study the viability of HPLC-FLD and HPLC-UV fingerprints

as sample chemical descriptors, the reproducibility of the method, and the

robustness of the chemometric results.

5. PLS-DA will be employed to study the samples' classification according to their

geographical origin and build a classification decision tree to calculate the

sensitivity, specificity, and accuracy of both methods.


5. EXPERIMENTAL SECTION

5.1. REAGENTS AND SOLUTIONS

The solvents used for extraction in the sample treatment, as well as to optimize and carry

out the chromatographic separation, were:

- Acetonitrile (ACN UHPLC Supergradient quality from Panreac)

- Formic acid (≥96% from Sigma-Aldrich)

- Methanol (99.9% from Panreac)

- Mili-Q water: Water was purified using an Elix 3 coupled to a Mili-Q system (Millipore,

Bedford, MA, USA) and filtered through a 0.22 µm nylon filter integrated into the Mili-Q

system.

5.2. INSTRUMENTATION AND CHROMATOGRAPHIC CONDITIONS

For the liquid chromatography separation, an Agilent 1100 Series HPLC instrument was

employed. It was equipped with a binary pump (G1312A), a vacuum degasser (G1379A), an

autosampler (G1367A), a diode array detector (G1315B) in series with a fluorescence detector

(G1321A), and a computer with the Agilent ChemStation software.

The chromatographic separation was carried out using a Kinetex core-shell C18 column

(100 mm x 4.6 mm i.d., 2.6 µm particle size) and guard column (2 mm × 4.6 mm i.d., 2.6 µm

particle size), both from Phenomenex (Torrance, CA, USA), and a mobile phase composed of

formic acid 0.1% (v/v) aqueous solution (solvent A) and acetonitrile (solvent B). Throughout the

chromatographic run, the flow rate was established at 0.5 mL·min-1, and the used gradient

elution is shown in Table 1. As seen, to obtain the best chromatographic separation, a

combination of isocratic and lineal gradients at 3 different acetonitrile concentrations was the

chosen option. Besides, an injection volume of 5 µL was set up for each sample.


Table 1. Gradient elution used in the chromatographic separation.

Time [min] Solvent B [%] Elution mode

0 - 2 40 Isocratic

2 - 3 40 - 80 Lineal

3 - 8 80 Isocratic

8 - 10 80 - 100 Lineal

10 - 12 100 Isocratic

12 - 13 100 - 40 Lineal

13 - 18 40 Isocratic

Regarding the fingerprint’s detection, 280 nm was chosen as the absorbance wavelength for

UV detection, while 310 and 380 nm were set as the excitation and emission wavelengths for

FLD acquisition, respectively.

5.3. SAMPLES AND SAMPLE TREATMENT

A total of 122 paprika samples purchased from markets in Spain, Czech Republic and

Hungary were analyzed. Figure 1 shows the number of samples from each region and the

relative percentage of flavors within each one. For example, out of the 45 samples from La

Vera, a third were hot flavors, so, 15 hot paprika samples from La Vera were studied.

Figure 1. Pie chart of the number of samples of the 5 studied regions (La Vera, Czech Republic, Mallorca,

Murcia and Hungary) and a 100% stacked bar chart with the respective flavors.

45

28

18

16

15

La Vera

Czech

Republic

Mallorca

Murcia

Hungary

0% 50% 100%

Hot Sweet Smoked sweet Bittersweet


Sample treatment was carried out following a previously described method [18]. Briefly, 0.3

g of paprika were weighed in a 15 mL polypropylene tube, and 3 mL of water:acetonitrile (20:80

v/v) were added. Then, the mixture was stirred for 1 min in a Vortex (Stuart, Stone, United

Kingdom), sonicated for 15 min (5510 Branson ultrasonic bath, Hampton, NH, USA), and

centrifuged 30 min at 4500 rpm (ROTANTA 460 HR Centrifuge, Hettich, Germany). Finally, the

resulting supernatant extract was filtered through a 0.22 μm nylon filter and preserved at 4ºC in

a 2 mL glass injection vial until analysis.

To minimize the influence of instrumental drifts in the subsequent chemometric models, all

samples were randomly analyzed. Moreover, a quality control (QC) sample, consisting of a

mixture prepared with 50 µL of each sample extract, as well as an extracting solvent blank,

were injected at the beginning and after every ten sample injections to control the repeatability

of the method and cross-contamination during sample sequence, respectively.

5.4. DATA ANALYSIS

Once the samples were analyzed, the raw data obtained with the Agilent ChemStation

software was exported to a spreadsheet using UniChrom™, from Seachrom Inc. (Lynnwood,

WA, USA). At this point, all the chromatographic fingerprints could be represented and Solo, a

stand-alone chemometrics software from Eigenvector Research (Manson, WA, USA), was used

for the PCA and PLS-DA calculations.

In this study, both PCA and PLS-DA were used. While non-supervised PCA was used to

observe the QC sample behavior and to study the samples’ structure, supervised PLS-DA was

employed for sample classificatory purposes. Indistinctly of the chemometric method used, the

construction of different data matrices was required. The X-data matrix, which was used either

in PCA and PLS-DA, consisted of the HPLC-UV (λ = 280 nm) or HPLC-FLD (λexc = 310 nm,

λem = 380 nm) chromatographic fingerprints obtained. Besides, in PLS-DA, the class

membership of each analyzed sample was defined in the Y-data matrix. In order to build the

PLS-DA models, the first significant minimum point of the cross-validation (CV) error from the

Venetian blind approach was chosen as the most appropriate number of LVs. Moreover, aiming

at the improvement of data quality, the chromatographic fingerprints were smoothed, baseline-

corrected, aligned, and autoscaled.


Considering the complexity of the studied issue, different PLS-DA models were

consecutively combined constituting a classification decision tree, using the hierarchical model

builder. The applicability of the chemometric method was then evaluated by external validation.

First, PLS-DA calibration (CAL) models were created using 60% of the paprika samples

(stratified random chosen). Instead, the remaining 40% was used as the validation (VAL) set.


6. RESULTS AND DISCUSSION

6.1. CHROMATOGRAPHIC SEPARATION

The objective of this work was to develop a HPLC-FLD method for the characterization,

classification, and authentication of paprika according to their region of origin. In order to obtain

the richest chromatographic fingerprints (i.e. with the maximum number of separated peaks),

one sample from La Vera was employed for the optimization of the chromatographic separation

in HPLC-FLD.

First, a mobile phase composed of formic acid 0.1% (v/v) aqueous solution and methanol

was used in a simple gradient, as seen in Figure 2A. After 0.5 min at 5% organic phase, a linear

gradient to 85% took place in 3.5 min and was followed by another linear gradient to 95% in 4

min and remained at that percentage one more min before returning to the initial conditions. The

obtained chromatographic fingerprint is observed in Figure 2B.

Figure 2. (A) First gradient elution tried in the chromatographic separation using methanol as solvent B and

(B) the respective HPLC-FLD fingerprint (λexc = 310 nm, λem = 380 nm) of a sample from la Vera.

Since the compounds were grouped in the same area where the methanol was 95%, it was

decided to use a less polar solvent with the purpose of having a better elution since the

chromatographic separation was carried out with a reversed-phase column. Even employing

0

7

14

0 4,5 9

8

0

0

4.5 9.0

16

Retention time [min]

FL

Inte

nsity

[r.f.u

. x 1

0E6]

0

20

40

60

80

100

0 1 4 8 9 10 15

Time [min]

Sol

vent

B [%

]

A B


acetonitrile as solvent B with a simple linear gradient from 5% to 95%, no remarkable change in

retention was observed and, for that reason, a different gradient elution program was proposed.

In the beginning, a linear increase from 40% to 85% was done in 1.5 min. After 5.5 min at 85%,

another increase to 100% was carried out in 2 min. The column was re-equilibrated after that

percentage remained one more minute. A big improvement was then obtained in the

chromatographic separation of paprika detected signals, especially in the retention time, but

with the aim of obtaining more resolved and separated peaks, the gradient elution mentioned in

Section 5.3 (Figure 3A) was chosen.

The chromatographic fingerprint obtained after extending the initial part with an isocratic

40% of solvent B (Figure 3B) had enough quality, at priori, for the objective of this work

(untargeted method), so, no more improvements in the chromatographic fingerprint separation

were tried. Since from 12 to 18 min the column was re-equilibrated at the initial conditions for

the next chromatographic injection, in this work, fingerprints only encompass from 0 to 12 min.

Figure 3. (A) Optimized gradient elution employed in the chromatographic separation using acetonitrile as

Solvent B and (B) the respective HPLC-FLD fingerprint (λexc = 310 nm, λem = 380 nm) of a sample from

la Vera.

6.2. CHROMATOGRAPHIC FINGERPRINTS

As previously mentioned, paprika samples from five different regions were analyzed by the

proposed HPLC-FLD method for classification purposes. In the end, both HPLC-UV (λ = 280

nm) and HPLC-FLD (λexc = 310 nm, λem = 380 nm) chromatographic fingerprints were studied.

0

20

40

60

80

100

2 3 8 10 120 13 18

Time [min]

Sol

vent

B [%

]

0

7

14

0 6 12

9

00

6 12

18


FL

Inte

nsity

[r.f.u

. x 1

0E6]

BA

0 6 12

8

0

16

8

0

16


6.2.1. HPLC-FLD chromatographic fingerprints

To begin with, an HPLC-FLD fingerprint of a random spicy sample from each region is

displayed in Figure 4.

Figure 4. HPLC-FLD chromatographic fingerprints (λexc = 310 nm, λem = 380 nm) for a random spicy paprika sample within each region.

At a first glance, HPLC-FLD fingerprints from La Vera and Mallorca can be differentiated

from those of the other regions without any problem. After the huge peak related to non-retained

compounds, they are characterized by a richer area (regarding detected bioactive compounds

and their relative abundances) from 6 to 12 min. HPLC-FLD fingerprints obtained for Czech

Republic, Murcia, and Hungary were also similar, in fact, the detected compounds also

appeared mainly in the same area and providing very similar peak signal distribution. Anyway,

peak intensity variations among all the regions can be easily spotted, hence, the obtained

0

8

16

0 6 12

8

00

6 12

16


FL

Inte

nsity

[r.f.u

. x 1

0E6]

0

8

16

0 6 12

8

00

6 12

16


FL

Inte

nsity

[r.f.u

. x 1

0E6]

8

00

6 12

16


FL

Inte

nsity

[r.f.u

. x 1

0E6]

0

8

16

0 6 12

8

00

6 12

16


FL

Inte

nsity

[r.f.u

. x 1

0E6]

3.2

00

6 12

6.4


FL

Inte

nsity

[r.f.u

. x 1

0E6]

La Vera

Mallorca

Czech Republic

HungaryMurcia

x2.5 zoom


HPLC-FLD fingerprints were employed as sample chemical descriptors for the chemometric

study.

6.2.2. HPLC-UV chromatographic fingerprints

The HPLC-UV fingerprints of the same samples above are shown in Figure 5.

Figure 5. HPLC-UV chromatographic fingerprints (λ = 280 nm) for a selected spicy paprika sample within each region.

On this occasion, all the chromatographic fingerprints seem to have the same detected

compounds, but with differences of relative peak intensities between regions. Again, as

differences between the relative intensities among the analyzed paprika samples are observed,

these HPLC-UV fingerprints were also evaluated as sample chemical descriptors for the

chemometric classification of the analyzed samples.

Retention time [min]6

00

6

12

12

Abs

orba

nce

(a.u

.x 1

0E5)


00

6

12

12

Abs

orba

nce

(a.u

. x 1

0E5)


00

6

12

12

Abs

orba

nce

(a.u

. x 1

0E5)


00

6

12

12

Abs

orba

nce

(a.u

. x 1

0E5)


00

6

12

12

Abs

orba

nce

(a.u

. x 1

0E5)La Vera

Mallorca

Czech Republic

HungaryMurcia


6.3. EXPLORATORY STUDIES BY PCA

In order to evaluate the usefulness, repeatability, and robustness of HPLC-FLD method for

paprika classification, a non-supervised exploratory PCA study was performed with the obtained

HPLC-FLD and HPLC-UV fingerprints. As mentioned in Section 5.4, raw data was pretreated to

reduce noise interferences, baseline drifts, and peak shifting. Furthermore, by decreasing

differences in the magnitude scale, the same influence was granted to each variable [33].

6.3.1. HPLC-FLD study by PCA

With this objective in mind, a data matrix (135 × 1669) consisting of the recorded

fluorescence intensity signals for each paprika sample and QCs was built, and subjected to

PCA. The Samples vs. QCs plot of scores of PC1 vs. PC2 is shown in Figure 6.

Figure 6. Scores plot of PC1 versus PC2 between QCs and paprika samples corresponding to HPLC-FLD fingerprints.

As it can be seen, QCs form a cluster, proving there were no experimental errors and the

good performance of the proposed methodology, so, chemometric results were considered

reliable. Thus, no correction was needed for the data matrix and an exploratory PCA

distinguishing among regions was done. Figure 7 shows the obtained plot of scores PC1 vs.

PC2.

-100 0 100-40

0

40

Scores on PC 1 (72.34%)

Sco

res

onP

C 2

(11

.35%

)

Samples

Quality controls

Samples

Quality controls


Figure 7. Scores plot of PC1 versus PC2 between regions and without QCs, corresponding to HPLC-FLD fingerprints.

As expected, La Vera and Mallorca can be easily distinguished from the others, probably

due to the main differences observed in their HPLC-FLD fingerprints as commented in section

6.2.1. On the other hand, Czech Republic and Hungary paprika samples, overlap with Murcia

samples, although samples tend to be grouped according to their production region. A peculiar

thing is that among Czech Republic and Mallorca there are 3 and 2 differentiated groups,

respectively, because this PCA has been able to differentiate between some sample flavors.

Even without using a supervised method, it seems that the HPLC-FLD fingerprints are good

enough to be used as sample chemical descriptors to address paprika classification and

authentication.

6.3.2. HPLC-UV study by PCA

A data matrix (135 × 1669) with absorbance signals of HPLC-UV fingerprints was built, and

the equivalent PCA scores plots as those previously described for HPLC-FLD are represented

in Figure 8.

-40

0

40

Sco

res

onP

C 2

(11

.43%

)

-100 0 100

Czech Republic

Murcia

Hungary

Mallorca

La Vera


Czech Republic

Murcia

Hungary

Mallorca

La Vera


Figure 8. Scores plot of PC1 versus PC2 corresponding to HPLC-UV fingerprints (A) between QCs and paprika samples (B) between paprika regions and without QCs.

QCs also form a cluster, as expected, showing the good performance of the proposed

methodologies. Regarding the grouping of samples according to their production region, HPLC-

UV seems to provide worse results, especially for some paprika samples produced in Mallorca,

which are now clustered close to Hungary paprika samples, which at the same time are

clustered close to samples from Murcia and Czech Republic. For now, the proposed HPLC-FLD

fingerprints seem to provide better results than the HPLC-UV ones. Nevertheless, as far as

classification concerns, nothing can be concluded because, as told, PCA is only a non-

supervised exploratory method.

6.4. CLASSIFICATION OF SAMPLES BY PLS-DA

With the aim of improving the PCA results and organize the classification decision tree, a

supervised classification chemometric method such as PLS-DA was used.

Samples

Quality controls

-100 0 100-40

0

40

Scores on PC 1 (63.55%)S

core

s on

PC

2 (

9.65

%)

-100 0 100-40

0

40


Sco

res

onP

C 2

(9.

60%

)

A B

Czech Republic

Murcia

Hungary

MallorcaLa Vera


6.4.1. HPLC-FLD classification by PLS-DA

As indicated in Section 5.4, a Y-data matrix (122 × 1670) was built removing the QCs and

including the class (region) of each sample in the X-matrix employed in HPLC-FLD PCA. Then,

a scores plot of LV1 vs. LV2 (five, the first minimum point of the CV error, was chosen as the

number of LVs) was represented. It is observed in Figure 9.

Figure 9. Scores plot of LV1 versus LV2 between regions, corresponding to HPLC-FLD fingerprints.

Unexpectedly, the PLS-DA scores plot shows no improvement regarding the exploratory

PCA. Most probably, the obtained HPLC-FLD fingerprints are sufficiently discriminating so that

an unsupervised method allows the correct discrimination without the need to specify the class.

To evaluate the classification capabilities of the proposed models, a classification plot of

sample vs. Y-predicted class for each region against the others was represented using the CAL

matrix (78 × 1670) explained in Section 5.4 as training set. The VAL matrix (48 × 1670) was

employed for the prediction results, which were expressed as the proportion of the studied

region samples that are correctly identified (sensitivity) and the proportion of correctly

recognized as others (specificity). Table 2 shows the results of the predictions for the five

regions, and Figure 10 serves as an example of the best and the worst obtained plots. In the

-40

0

40

Sco

res

onLV

2 (X

: 11.

43%

/ Y

: 19.

41%

)

-100 0 100

Czech Republic

Murcia

Hungary

Mallorca

La Vera

Scores on LV 1 (X: 72.68% / Y: 25.70%)

Czech Republic

Murcia

Hungary

Mallorca

La Vera


latter ones, the red mark represents the classification threshold line. Above this, samples are

assigned to a predefined class and below to the opposite class.

Table 2. Sensitivity and specificity results for each region prediction in the first sequence.

Region Sensitivity [%] Specificity [%]

Czech Republic 100.0 95.2

Hungary 75.0 100.0

Mallorca 100.0 100.0

Murcia 100.0 81.0

La Vera 100.0 100.0

Figure 10. Sample versus Y predicted 1 scores plot with HPLC-FLD calibration and validation matrix for (A)

La Vera vs. other regions and (B) Murcia vs. other regions.

Then, to build a classification decision tree, the region with the best prediction results was

removed from the matrices. In case of a tie, the region with more replicates was considered

better. In this case, La Vera prediction model was saved and the 45 samples were eliminated

(27 from CAL matrix and 18 from VAL matrix), then, another set of predictions was done with

the remaining regions. The process was repeated till only two regions remained. Table 3 and

Table 4 show the obtained results for the last sets.

20 40 60 80 100 120-0.2

0

0.2

0.4

0.6

0.8

1.0

1.2

Sample

Y P

redi

cted

1 (L

a V

era)

Training set Validation set

La Vera La Vera

Others Others

A

Others Others


Murcia Murcia

20 40 60 80 100 120-0.2

-0.1

0

0.1

0.2

0.3

0.4

0.5

Sample

Y P

redi

cted

1 (M

urci

a)B


Table 3. Sensitivity and specificity results for each region prediction in the second sequence.


Czech Republic 100.0 100

Hungary 91.6 100.0


Murcia 100.0 75.0

Table 4. Sensitivity and specificity results for each region prediction in the third sequence.



Hungary 83.3 100.0

Murcia 83.3 77.8

Mallorca and Czech Republic prediction models for the second and third sequence were

saved respectively, as well as the model obtained from comparing Hungary and Murcia. Figure

11 illustrates the Y-predicted plots of the chosen models for the classification decision tree.

Figure 11. Sample versus Y predicted 1 scores plot with HPLC-FLD calibration and validation matrix for (A)

Mallorca vs. other regions (without La Vera), (B) Czech Republic vs. other regions (without La vera and

Mallorca), and (C) Hungary vs. Murcia.

10 20 30 40 50 60 70-0.2

0

0.2

0.4

0.6

0.8

1.0

Sample

Y P

redi

cted

1 (M

allo

rca)

10 20 30 40 50 60-0.2

0

0.2

0.4

0.6

0.8

1

1.2

Sample

Y P

redi

cted

1 (C

zech

Rep

ublic

)

10 20 30 40-0.2

0

0.2

0.4

0.6

0.8

1

1.2

Sample

Y P

redi

cted

1 (H

unga

ry)


Hungary Hungary

Training set

Others


Others Others

Mallorca Mallorca Czech Republic

Validation set

Others

Czech Republic

Murcia Murcia

A B C


6.4.2. HPLC-UV classification by PLS-DA

The classification of paprika samples by employing HPLC-UV fingerprints as chemical

descriptors was also performed following the same process as previously described with HPLC-

FLD fingerprints. In Figure 12 two PLS-DA scores plots (choosing six as the most appropriate

number of LVs) are represented.

Figure 12. (A) Scores plot of LV1 versus LV2 and (B) LV1 versus LV3, both between regions and

corresponding to HPLC-UV fingerprints.

In this case, and as expected, a great improvement over the PCA scores plot is appreciated.

Also, since more than one comparison between LVs give a good region separation (Hungary is

completely differentiated in LV1 vs. LV3 scores plot), it could be said that HPLC-UV fingerprints

seem to provide better classifications than HPLC-FLD fingerprints. To make this sure, the

classification capabilities using the HPLC-UV data was also carried out.

As seen in Figure 13, the Y-predicted plots seem similar to the above, for that reason, the

same regions in the same order as HPLC-FLD were used as a model for the HPLC-UV to build

a classification decision tree.

-100 0 100

-20

0

20

Scores on LV 1 (X: 63.54% / Y: 25.89%)

Sco

res

onLV

3 (

X: 3

.07%

/ Y

: 16.

37%

)

-100 0 100-40

0

40

Scores on LV 1 (X: 63.54% / Y: 25.89%)

Sco

res

onLV

2 (

X: 9

.16%

/ Y

: 16.

83%

)

A B

Czech Republic MurciaHungary MallorcaLa Vera


Figure 13. Sample versus Y predicted 1 scores plot with HPLC-UV calibration and validation matrix for (A)

La Vera vs. other regions (first set) and (B) Hungary vs. Murcia (last set).

6.5. VALIDATION RESULTS

As previously commented, once the best models were chosen with the PLS-DA

classification method, the decision tree was created with the help of the hierarchical model

builder, using as rule the saved CAL models. Figure 14 illustrates the classification decision tree

employed for the validation of the HPLC-FLD and HPLC-UV methods.

Figure 14. Classification decision tree employed for the categorization of the validation matrix of HPLC-FLD and HPLC-UV.

10 20 30 40-0.4

-0.2

0

0.2

0.4

0.6

0.8

1.0

1.2


Hungary Hungary

Murcia Murcia


La Vera

Others

La Vera

Others

20 40 60 80 100 120

0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

Sample

Y P

redi

cted

1 (L

a V

era)

Sample

Y P

redi

cted

1 (H

unga

ry)

A B

String

String

String

String

String

String

String

String

String

Rule

Rule

Rule

Rule

others

others

others

La Vera

Mallorca

Hungary

Murcia

Czech R.

= Misclassified

= Misclassified

= Misclassified

= Misclassified

= La Vera

= Mallorca

= Czech Republic

= Hungary

= Murcia


The obtained results after analyzing the HPLC-FLD VAL matrix with the built tree are

observed in Table 5. As expected, sensitivity and specificity for La Vera and Mallorca were

100%. The classification PLS-DA was always perfect and, even in the exploratory HPLC-FLD

PCA, these regions were completely differentiated from the others. Czech Republic, Murcia and

Hungary had similar fingerprints and appeared grouped in the PCA and PLS-DA classification,

so, the confusion between these samples was not strange. However, the prediction capabilities

of the proposed HPLC-FLD fingerprinting method (specificity) is good, with values always higher

than 97.6%.

Table 5: Prediction results for HPLC-FLD.



Hungary 83.3 100.0


Murcia 100.0 97.6

La Vera 100.0 100.0

Instead, the classification results with HPLC-UV fingerprints are shown in Table 6. Similar

values as before were obtained, for the same reasons. Now, two samples from Hungary (not the

same ones for both methods) were classified as Murcia. At first glance, it was strange because

one of the PLS-DA scores plot distinguished Hungary from the other regions perfectly, but later,

in the prediction validation (Figure 13B), it is clearly shown that these two samples were

misinterpreted as Murcia ones. Again, the prediction capabilities of the proposed HPLC-UV

fingerprinting method (specificity) is good, with values always higher than 95.2%.

Table 6: Prediction results for HPLC-UV.



Hungary 83.3 100.0


Murcia 100.0 95.2

La Vera 100.0 100.0


In both situations, only two samples from Hungary were incorrectly classified. Therefore, a

good accuracy (95.8%) for both HPLC-FLD and HPLC-UV fingerprinting methods in the

classification of paprika samples was achieved, demonstrating the feasibility of the proposed

fingerprinting methodologies to address the characterization, classification, and authentication

of paprika samples regarding their geographical indication.


7. CONCLUSIONS

In this work, non-targeted HPLC-FLD (λexc = 310 nm, λem = 380 nm) and HPLC-UV (λ =

280 nm) methods to achieve characterization, classification and authentication of paprika

regarding their geographical indication were developed.

The optimized gradient elution employed in the chromatographic separation had enough

quality to make the fingerprints obtained from both detection systems under study viable sample

chemical descriptors to address sample classification by chemometrics. Also, the proposed

methodologies were reproducible and robust according to the QCs behavior.

In reference to the results obtained in the classification of paprika samples, HPLC-FLD and

HPLC-UV fingerprinting methods seem to have great accuracy, with 95.8% in both cases.

However, the prediction capabilities of the HPLC-FLD fingerprints (specificity > 97.6%) are a

little higher than the HPLC-UV ones (specificity > 95.2%). Even so, in the case of prioritizing

one, HPLC-UV fingerprinting would be chosen since it is cheaper, and it can be found in many

more laboratories than HPLC-FLD. Even in the scenario where the objective is to differentiate

between paprika tastes from the same region, HPLC-UV detection would be selected. Although

this data has not been discussed in the present work, preliminary results seem to indicate that

the classification of paprika regarding their flavor by using HPLC-UV fingerprints has much

better accuracy than with the HPLC-FLD ones.

Future work will focus on studying the adulteration of paprika with samples from another

geographical origin by partial least squares regression.


8. REFERENCES

[1] W. Verbeke, “Agriculture and the food industry in the information age,” Eur. Rev. Agric.

Econ., vol. 32, no. 3, pp. 347–368, 2005.

[2] G. P. Danezis, A. S. Tsagkaris, F. Camin, V. Brusic, and C. A. Georgiou, “Food

authentication: Techniques, trends & emerging approaches,” TrAC - Trends Anal.

Chem., vol. 85, pp. 123–132, 2016.

[3] J. C. Moore, J. Spink, and M. Lipp, “Development and Application of a Database of

Food Ingredient Fraud and Economically Motivated Adulteration from 1980 to 2010,”

Journal of Food Science. 2012.

[4] N. Carranco, M. Farrés-Cebrián, J. Saurina, and O. Núñez, “Authentication and

quantitation of fraud in extra virgin olive oils based on HPLC-UV fingerprinting and

multivariate calibration,” Foods, vol. 7, no. 4, 2018.

[5] C. V. Di Anibal, I. Ruisánchez, and M. P. Callao, “High-resolution 1H Nuclear Magnetic

Resonance spectrometry combined with chemometric treatment to identify adulteration

of culinary spices with Sudan dyes,” Food Chem., vol. 124, no. 3, pp. 1139–1145,

2011.

[6] Y. Liu, E. E. D. Todd, Q. Zhang, J. R. Shi, and X. J. Liu, “Recent developments in the

detection of melamine,” Journal of Zhejiang University: Science B. 2012.

[7] S. Medina, J. A. Pereira, P. Silva, R. Perestrelo, and J. S. Câmara, “Food fingerprints –

A valuable tool to monitor food authenticity and safety,” Food Chem., vol. 278, no. July

2018, pp. 144–162, 2019.

[8] Consejo de la union europea, “Reglamento (CE) N 510/2006 del consejo de 20 de

marzo de 2006 sobre la proteccion de las indicaciones geograficas y de las

denominaciones de origen en los productos agricolas y alimenticios,” D. Of. la union

Eur., no. 10, pp. 12–25, 2006.

[9] N. Serrano et al., “Characterization and classification of Spanish paprika (Capsicum

annuum L.) by liquid chromatography coupled to electrochemical detection with screen-

printed carbon-based nanomaterials electrodes,” Talanta, vol. 189, no. July, pp. 296–

301, 2018.

[10] I. Bata-Vidács et al., “Investigation of regional differences of the dominant microflora of

spice paprika by molecular methods,” Food Control, vol. 83, pp. 109–117, 2018.

[11] A. S. Antonio, L. S. M. Wiedemann, and V. F. Veiga Junior, “The genus: Capsicum: a

phytochemical review of bioactive secondary metabolites,” RSC Adv., vol. 8, no. 45,

pp. 25767–25784, 2018.


[12] S. Barbosa, G. Campmajó, J. Saurina, L. Puignou, and O. Núñez, “Determination of

Phenolic Compounds in Paprika by Ultrahigh Performance Liquid Chromatography-

Tandem Mass Spectrometry: Application to Product Designation of Origin

Authentication by Chemometrics,” J. Agric. Food Chem., vol. 68, no. 2, pp. 591–602,

2020.

[13] S. Barbosa, J. Saurina, L. Puignou, and O. Núñez, “Classification and authentication of

paprika by UHPLC-HRMS fingerprinting and multivariate calibration methods (PCA and

PLS-DA),” Foods, vol. 9, no. 4, 2020.

[14] S. Kim et al., “Genome sequence of the hot pepper provides insights into the evolution

of pungency in Capsicum species,” Nat. Genet., vol. 46, no. 3, pp. 270–278, 2014.

[15] Y. Fiamegos, C. Dumitrascu, S. Papoci, and M. B. de la Calle, “Authentication of PDO

paprika powder (Pimentón de la Vera) by multivariate analysis of the elemental

fingerprint determined by ED-XRF. A feasibility study,” Food Control, vol. 120, p.

107496, 2021.

[16] M. Ghasemnezhad, M. Sherafati, and G. A. Payvast, “Variation in phenolic compounds,

ascorbic acid and antioxidant activity of five coloured bell pepper (Capsicum annum)

fruits at two different harvest times,” J. Funct. Foods, vol. 3, no. 1, pp. 44–49, 2011.

[17] J. Mateo, M. Aguirrezábal, C. Domínguez, and J. M. Zumalacárregui, “Volatile

Compounds in Spanish Paprika,” J. Food Compos. Anal., vol. 10, no. 3, pp. 225–232,

1997.

[18] X. Cetó et al., “Determination of HPLC-UV fingerprints of spanish paprika (Capsicum

annuum L.) for its classification by linear discriminant analysis,” Sensors (Switzerland),

vol. 18, no. 12, 2018.

[19] O. Fayos et al., “Assessment of Capsaicinoid and Capsinoid Accumulation Patterns

during Fruit Development in Three Chili Pepper Genotypes (Capsicum spp.) Carrying

Pun1 and pAMT Alleles Related to Pungency,” J. Agric. Food Chem., vol. 67, no. 44,

pp. 12219–12227, 2019.

[20] L. M. Reid, C. P. O’Donnell, and G. Downey, “Recent technological advances for the

determination of food authenticity,” Trends Food Sci. Technol., vol. 17, no. 7, pp. 344–

353, 2006.

[21] O. Núñez and P. Lucci, “Application of liquid chromatography in food analysis,” Foods,

vol. 9, no. 9, pp. 1–5, 2020.

[22] A. Palacios-Morillo, J. M. Jurado, A. Alcázar, and F. Pablos, “Differentiation of Spanish

paprika from Protected Designation of Origin based on color measurements and

pattern recognition,” Food Control, vol. 62, pp. 243–249, 2016.

[23] O. Monago-Maraña, C. E. Eskildsen, T. Galeano-Díaz, A. Muñoz de la Peña, and J. P.

Wold, “Untargeted classification for paprika powder authentication using visible – Near

infrared spectroscopy (VIS-NIRS),” Food Control, vol. 121, no. June 2020, 2021.

[24] B. Horn, S. Esslinger, M. Pfister, C. Fauhl-Hassek, and J. Riedl, “Non-targeted

detection of paprika adulteration using mid-infrared spectroscopy and one-class


classification – Is it data preprocessing that makes the performance?,” Food Chem.,

vol. 257, no. March, pp. 112–119, 2018.

[25] S. Lohumi, H. Lee, M. S. Kim, J. Qin, and B. K. Cho, “Raman imaging for the detection

of adulterants in paprika powder: A comparison of data analysis methods,” Appl. Sci.,

vol. 8, no. 4, 2018.

[26] Y. Hu, S. Wang, S. Wang, and X. Lu, “Application of nuclear magnetic resonance

spectroscopy in food adulteration determination: The example of Sudan dye i in paprika

powder,” Sci. Rep., vol. 7, no. 1, pp. 1–9, 2017.

[27] X. Cetó, C. Sánchez, N. Serrano, J. M. Díaz-Cruz, and O. Núñez, “Authentication of

paprika using HPLC-UV fingerprints,” Lwt, vol. 124, no. August 2019, 2020.

[28] N. Z. Ballin and K. H. Laursen, “To target or not to target? Definitions and

nomenclature for targeted versus non-targeted analytical food authentication,” Trends

Food Sci. Technol., vol. 86, no. September 2018, pp. 537–543, 2019.

[29] L. Cuadros-Rodríguez, C. Ruiz-Samblás, L. Valverde-Som, E. Pérez-Castaño, and A.

González-Casado, “Chromatographic fingerprinting: An innovative approach for food

‘identitation’ and food authentication - A tutorial,” Anal. Chim. Acta, vol. 909, pp. 9–23,

2016.

[30] S. Esslinger, J. Riedl, and C. Fauhl-Hassek, “Potential and limitations of non-targeted

fingerprinting for authentication of food in official control,” Food Res. Int., vol. 60, pp.

189–204, 2014.

[31] D. Granato et al., “Trends in Chemometrics: Food Authentication, Microbiology, and

Effects of Processing,” Compr. Rev. Food Sci. Food Saf., vol. 17, no. 3, pp. 663–677,

2018.

[32] J. Trygg, E. Holmes, and T. Lundstedt, “Chemometrics in metabonomics,” Journal of

Proteome Research. 2007.

[33] G. Campmajó, L. Cayero, J. Saurina, and O. Núñez, “Classification of Hen Eggs by

HPLC-UV fingerprinting and chemometric methods,” Foods, vol. 8, no. 8, pp. 1–10,

2019.


9. ACRONYMS

ACN: Acetonitrile

a.u.: Arbitrary Units

CAL: Calibration

CV: Cross-Validation

EC: Council Regulation

EFSA: European Food Safety Authority

FL: Fluorescence

HPLC: High-Performance Liquid Chromatography

HPLC-FLD: High-Performance Liquid Chromatography with fluorescence detection

HPLC-UV: High-Performance Liquid Chromatography with ultraviolet detection

i.d.: Internal Diameter

ISO: International Standardization Organization

LC-HRMS: Liquid Chromatography coupled to High-Resolution Mass Spectrometry

LC-MS: Liquid Chromatography coupled to Mass Spectrometry

LC-UV: Liquid Chromatography with ultraviolet detection

LV: Latent Variable

NMR: Nuclear Magnetic Resonance

PC: Principal Component

PCA: Principal Component Analysis

PDO: Protected Designation of Origin

PLS-DA: Partial Least Squares - Discriminant Analysis

QC: Quality Control

r.f.u.: Relative Fluorescence Units


UHPLC: Ultra High-Performance Liquid Chromatography

UV-VIS: Ultraviolet-Visible

VAL: Validation

Treball Final de Graudiposit.ub.edu/dspace/bitstream/2445/175436/1/TFG_QU...uno de los problemas....

Documents

Transcript of Treball Final de Graudiposit.ub.edu/dspace/bitstream/2445/175436/1/TFG_QU...uno de los problemas....